Bull. Mater. Sci., Vol. 23, No. 5, October 2000, pp. 405–411. © Indian Academy of Sciences.

A model for two-step ageing

K T KASHYAP*, C RAMACHANDRA, B CHATTERJI and S LELE†
Central Materials and Processes Laboratory, Foundry & Forge Division, Hindustan Aeronautics Limited,
Bangalore 560 017, India
*Department of Metallurgical Engineering, Banaras Hindu University, Varanasi 221 005, India

MS received 24 January 2000; revised 4 May 2000

Abstract. In commercial practice, two-step ageing is commonly used in Al–Zn–Mg alloys to produce a fine
dispersion of η ′ precipitates to accentuate the mechanical properties and resistance to stress corrosion cracking.
While this is true in Al–Zn–Mg alloys, two-step ageing leads to inferior properties in Al–Mg–Si alloys. This
controversial behaviour in different alloys can be explained by Pashley’s Kinetic model. Pashley’s model
addresses the stability of clusters after two-step ageing. In the development of the model, the surface energy
term between cluster and matrix is taken into account while the coherency strains between the cluster and
matrix are not considered. In the present work, a model is developed which takes into account the coherency
strains between cluster and matrix and defines a new stability criterion, inclusive of strain energy term.
Experiments were done on AA 7010 aluminium alloy by carrying out a two-step ageing treatment and the
results fit the new stability criterion. Thus it is found that the new model for two-step ageing is verified in the
case of Al–Zn–Mg alloy.

Keywords. Two-step ageing; delayed ageing; clusters; coherency strains.

1. Introduction perature and this leads to a coarse distribution of β ′ pre-

In Al–Zn–Mg alloys, preageing at a lower temperature
after solutionizing, followed by a higher temperature arti-
ficial ageing is known to give a fine dispersion of η′ pre-
cipitates (Lorimer and Nicholson 1966) and thus enhance
the mechanical properties. While this is true in Al–Zn–Mg
alloys, in Al–Mg–Si alloys, room temperature storage
after solutionizing, followed by high temperature artificial
ageing leads to a coarse precipitate ( β ′) structure (Pashley
et al 1966). This phenomenon is termed delayed ageing
wherein the delay after solutionizing before ageing leads
to inferior properties (Ghate et al 1984; Murali et al
1993).
Pashley et al have put forth a kinetic model to describe
the stability of clusters after two-step ageing, which quali-
tatively explains the controversy between Al–Zn–Mg and
Al–Mg–Si alloys. In Pashley’s kinetic model, a parameter
K is defined which is proportional to the surface energy
between the cluster and matrix which determines the sta-
bility of distribution of clusters in two-step ageing. In the
case of Al–Zn–Mg alloys, fine scale seeding takes place
at a lower temperature ageing which gets stabilized at
higher temperature ageing resulting in a fine dispersion of
η′ precipitates. However in Al–Mg–Si alloys only the
coarse clusters are stabilized at the artificial ageing tem-
*Author for correspondence
cipitates.
While Pashley’s kinetic model addresses the surface
energy between the cluster and the matrix, the elastic cohe-
rency strains between them are not considered. Therefore
it is the objective of the present work to include cohe-
rency strains between the cluster and matrix in the deve-
lopment of the Pashley’s kinetic model (extended Pashley’s
model) and verify the same.
2. The model
In the present model, elastic strains between the cluster
and matrix are considered. The elastic strains are intro-
duced via the classical definition of Gibbs free energy.
Thus the Gibbs–Thomson relationship is modified in that
not only the curvature effect is considered in terms of the
partial molar free energy but also elastic strain energy is
included. The Gibbs–Thomson relationship is the one that
describes the solubility as a function of curvature of an
interface. Here only the surface energy and curvature
have an influence on solubility. In the modified Gibbs–
Thomson relationship not only the curvature and surface
energy have an influence on the solubility but also the
elastic strain energy. Using the modified Gibbs–Thomson
relationship, the criterion for the stability of the clusters is
derived for two-step ageing. Further, it is also assumed
that the elastic modulus of the cluster is greater than that
405
406 K T Kashyap et al

of the matrix and therefore the coherency strains are acco- Vm Eå2
mmodated in the matrix. G i, r − G i, ∞ = Vm ( Pr − P∞ ) + , (8)
4
It is known from classical thermodynamics (Fast 1970),
that Gibbs free energy can be expressed as Vm Eε 2
= Vm ∆P + , (8a)
4
G = U − TS + ∑ XY , (1)
where, Gi,r is the partial molar free energy of a curved
where G is the Gibbs free energy, U, the internal energy, interface of radius ‘r’, Gi,∞ the partial molar free energy of
S, entropy, T, absolute temperature, ∑ XY the work done a flat interface and Pr, P∞ are pressures across a curved
due to elastic strains, thus ∑ XY can be expressed as two interface and a flat interface respectively.
terms viz. volume work (PdV) and work done due to ela- It is known from classical thermodynamics (Fast 1970)
stic strains, thus that
 å
G = U − TS + PV + Vm ó   , (2)  ai, r 
2 G i, r − G i ,∞ = RT ln  . (9)
 ai, 
Vm is molar volume, σ, stress in the matrix which is  ∞ 
elastic.
Differentiating (2), we have Combining (9) and (8a), where ai,r, ai,∞ are the activity of
the solute in a curved interface and flat interface, respec-
Vm V
dG = dU − TdS − SdT + PdV + VdP + ó då + m å dó . tively
2 2
(3)  ai, r    Vm Eå2
RT ln   = V m ã 1 + 1
 + .
 ai,   r
Assuming reversibility, the term dU can be expressed as  ∞   1 r2  4

Vm Since
dU = TdS − PdV − ó då . (4)
2  1 1 
∆P = γ  +  ,
Substituting (4) in (3)
 r1 r2 
Vm where, γ is the surface energy and r1, r2, are the principal
dG = VdP + å dó − SdT .
2 radii of curvature (Porter and Easterling 1992).
Assuming ideal solution behaviour,
At constant temperature (dT = 0),
a i = X i,
V
dG = VdP + m å dó. (5)
2 where Xi is the concentration,

Assuming the elastic strains to be small, i.e. Hooke’s law

 X i,r    2
is valid, RT ln   = Vm ã 1 + 1  + Vm Eå . (10)
X   r 
 i,∞   1 r2  4
σ = Eε ,

where E is the Young’s modulus, Assuming spherical clusters (it is believed that the physi-
cal reasoning behind the model is not altered because of
dσ = Edε. (6) the cluster shape)
Substituting (6) in (5),
 X i, r   Vm γ ( 2 ) Vm Eε 2 
ln  = + ,
Vm  X i, ∞   RTr 4 RT 
dG = VdP + εEdε .    
2
  V 2γ Vm Eε 2 
Expressing the above relationship in terms of partial X i,r = X i,∞  exp  m +  . (11)
quantities,   RTr 4 RT 
  
Vm
dGi = Vm dP + åEdε . (7) This is the modified Gibbs–Thomson relationship where
2
not only curvature has an effect on the solubility but also
Integrating (7) for a curved interface, strains between the cluster and matrix.
 A model for two-step ageing 407

In the original Pashley’s model (Pashley et al 1967), Vm 2γ

the stability of the clusters is dictated by the condition K= ,
RT
that the rate of arrival of solute to the cluster should
be greater than the rate of departure of solute from the (14) becomes
cluster.
Pashley et al (1967) have given the condition to be  A
r ln   S > K . (15)
B
AX∞S > BX, (12)
If it is assumed that no excess vacancies are present,
where, A is the coefficient which determines the rate of A = B, (15) becomes
arrival of solute to the clusters which is proportional to
the diffusion coefficient and excess vacancies. S is the r ln S > K. (16)
supersaturation = W(1 – α)/X∞, W, being the weight of the
solute/unit volume, α the fraction of solute in the clusters. This is the stability criterion in the original Pashley’s
X∞, concentration of the solute in equilibrium with a flat model.
interface, X, concentration of the solute in equilibrium Furthermore, assuming spherical clusters, mass balance
with a curved interface and B, the kinetic coefficient. of the solute leads to the following
The Gibbs–Thomson relationship without elastic
strains is 4
 V 2γ 
∑ 3 π r 3ρnr = αW , (17)
X i,r = X i,∞ exp  m  . (13)
 RTr 
where, ρ is the density of the cluster species, nr, the num-
Substituting (13) in (12) for X, Pashley arrived at the sta- ber of cluster per unit volume.
bility criterion, Figure 1 shows these relationships schematically, with a
delay at a lower temperature (T1), α increases and super-
 V 2γ  saturation in the matrix decreases and when α > αr, the
AX ∞ > BX i, ∞ exp  m  , clusters are stabilized at the artificial ageing tempera-
 RTr 
ture (T2).
Figure 2 shows the effect of size distribution of clusters
 A  V 2γ 
i.e. r ln  S >  m  . (14) on the stability after two step-ageing. At a lower tempera-
 
B  RT  ture (T1), clustering will take place resulting in a size
distribution of clusters. The size distribution can be
Letting, assumed to be gaussian and accordingly the variation of r

Figure 1. Schematic illustration of stability of clusters in two- Figure 2. Schematic illustration of stability of clusters with
step ageing. size distribution in the original model.
408 K T Kashyap et al

with α can be represented as shown in figure 2. In a simi-  A

lar way the term r ln S is also shown in figure 2. It is seen r ln   S > K + Qr. (18)
B
that when α > αrmax, only the coarse clusters are stabilized
at the artificial ageing temperature (T2) after a delay at the
Equation (18) is thus the general stability criterion for
lower temperature (T1). This effect explains the behaviour
clusters which introduce elastic strains in the matrix.
of Al–Mg–Si alloys which show delayed ageing (Murali
If it is assumed that there are no excess vacancies,
et al 1993) where room temperature storage (T1) gives
A = B, then (18) becomes
rise to a coarse precipitate structure after artificial age-
ing (T2).
Now, considering the elastic strains between the cluster r ln S > K + Qr. (19)
and the matrix, the modified Gibbs–Thomson relationship
(11), Pashley’s criterion (AX∞S > BX (i.e. (12)) for stabi- Figure 3 shows the stability diagram for clusters after two-
lity on clusters after two-step ageing can be expressed as step ageing with K and K + Qr. It is clearly shown in the
figure that the clusters with α1,r get stabilized when no
 V 2γ Vm Eε 2  strain is considered in the matrix. On the other hand,
AX ∞ S > BX i, ∞ exp  m + .
 RTr 4 RT  when elastic strains are considered, a much longer delay
  representing a situation like α2, (where α2, r >> α1, r) is
By substituting (11) in (12) required for the cluster to stabilize. Thus it can be stated
that clusters will find it more difficult to stabilize with the
 V 2γ V Eε 2  introduction of coherency strains.
 A
ln   S >  m + m , Figure 4 shows the effect of size distribution of clusters
B  RTr 4 RT 
  with increase in delay at T1 (lower temperature) corres-
ponding to increased α. It can be seen from figure 4 that
 A  V 2γ V Eε 2  only the large sized clusters are stabilized at αr,max in the
r ln   S >  m + r m .
B  RT 4 RT  original Pashley’s model whereas with coherency strains
 
such clusters are also destabilized. In the original model,
Letting such behaviour would have led to delayed ageing wherein
a coarse precipitate structure is evolved upon delay. But
Vm 2 r Vm Eε 2 with large coherency strains (large Q) delayed ageing is
K= , Q= ,
RT 4 RT not expected.

Figure 3. Schematic illustration of stability of clusters con- Figure 4. Schematic illustration of stability of clusters with
sidering elastic strains. size distribution considering elastic strains.
 A model for two-step ageing 409

3. Experimental Therefore,

In order to test the extended Pashley’s model, alloy AA 4G  1 + v 

ε=  δ. (20)
7010 was selected for delayed ageing studies. The compo- E  1− v 
sition of AA7010 is given in table 1.
Samples were sectioned and heat treated AA 7010 sam- It is known that,
ples were solutionized at 748 K for 50 min and water
E
quenched. The samples were delayed for 0, 2, 4, 20 and G= = 26 ⋅12 GPa for aluminium,
68 h respectively and then artificially aged at 443 K for 2(1 + v)
24 h. Samples were machined and tested for mechanical
E = 70 Gpa,
properties in an universal testing machine.
ν = 0⋅34.
4. Results and discussion
Table 3 shows the misfit strain calculations for intermedi-
ate precipitate based on the orientation relationship with
Table 2 shows yield strength as a function of delay in arti-
the matrix for several aluminium alloys. From the values
ficial ageing for AA7010 alloy. There is no substantial
of misfit strain (δ), the corresponding elastic strains (ε)
drop in the tensile strength with delay in ageing. The per-
have been evaluated using the relationship (20) and are
centage drop in the tensile strength is only 7⋅42% with
recorded in table 3. It can be seen that the misfit strain
20 h delay whereas in Al–Mg–Si alloy (Murali et al
parameter for Al–Mg–Si alloy is very small whereas in
1993), the percentage drop in tensile strength is 15%.
Al–Zn–Mg alloy, it is very large. Similarly elastic strain
Thus it can be seen that delayed ageing is more severe in
‘ε ’ is large for Al–Zn–Mg and small for Al–Mg–Si alloy.
Al–Mg–Si alloy.
It should be noted from recent literature (Maloney et al
To rationalize the above results, the following analysis
1999; Murayama and Hono 1999), that the chemistry of
was carried out. The elastic strain free energy is given by
the GP zones by atom probe and field ion microscopy in
Eshelby (Cahn et al 1991) as,
Al–Zn–Mg and Al–Mg–Si alloys, respectively is deter-
mined. The composition of the GP zone in Al–Mg–Si
 1+ v  2
∆Gel = 2G  δ , alloys is 50 at% Mg, 50 at% Si. The composition of GP
 1− v  zone in Al–Zn–Mg alloy is given by 8⋅7 at% Zn, 6⋅3 at%
Mg giving a ratio of Zn : Mg of 1⋅4 : 1.
where, G is the shear modulus, ν, the Poisson ratio.

a ppt − a matrix Table 2. Effect of delay in ageing on the tensile

δ, the misfit strain parameter = , strength of AA 7010 alloys.
a matrix
Alloy 7010 LM 25
appt is the lattice parameter of the precipitate, amatrix, the (Al–Zn–Mg) (Al–Mg–Si)
lattice parameter of the matrix. Delay
To relate elastic strain in the matrix due to coherency YS % Red. YS % Red.
(h)
(Mpa) (UTS) (Mpa) (UTS)
strains between precipitate and matrix
0 512 – 230 –
1+ v  2 20 474 7⋅42 195 15⋅2
1
σε = ∆Gel = 2G  δ .
 1− v 
2

1+ v  2 Table 3. Misfit strain values.

1
Eε 2 = 2G  δ .
 1− v 
2
Alloy Al–Zn–Mg Al–Mg–Si Al–Cu–Mg

Intermediate η′ β′ S′
Table 1. Compositions of AA 7010 and LM25. precipitate
Element Orientation (001)η′ (001) β (210)S
relationship | |(111)α | |(100)α | |(100) S
Zn Mg Li Cu Zr Si Fe Al Misfit (δ ) % –23⋅75 –7⋅4 × 10–3 –0⋅1216
Elastic strain (ε) –0⋅4134 –0⋅128 × 10–3 –2⋅1 × 10–3
AA 7010
(wt %)
6⋅2 2⋅5 – 1⋅5 – – – Bal η′(MgZn2) hexagonal a = 4⋅96 Å c = 8⋅68 Å
β ′(Mg2Si) hexagonal a = 7⋅05 Å c = 4⋅05 Å
LM 25
(wt %) – 0⋅3 – – – 7⋅0 0⋅5 Bal S′(Al2CuMg) orthorhombic a = 4⋅04 Å b = 9⋅25 Å
c = 7⋅18 Å
410 K T Kashyap et al

Table 4. Misfit strain values (GP zones).

Alloy Al–Mg–Si Al–Zn–Mg

3
Volume per atom 24⋅09 Å 31⋅136 Å3
Volume per atom (matrix) 24⋅60 Å3 29⋅520 Å3
Misfit δ 0⋅02073 –0⋅0597

Taking atomic ratio of Al, Mg, Si, Zn, the volume per
atom of the GP zones is calculated. Table 4 shows results
of the calculation. It is seen the misfit strain (diffe-
rence between volume/atom of zones and matrix) is low
(0⋅02073) for Al–Mg–Si alloy whereas it is high for Al–
Zn–Mg alloy (– 0⋅0547). Using (20) the elastic strain is
also calculated. It is also shown in table 4.
From the model, the stability criterion for clusters after
two-step ageing is given by the following equation

r ln S > K + Qr, (19)

where, Figure 5. Schematic illustration of stability situation for

Al–Mg–Si and Al–Zn–Mg alloys.

2γ Vm Eε 2
K= , Q= ,
RT 4 RT
Q is a dimensionless quantity whereas K has units of
length.
Assuming, Vm = 10–5 m3/mole, γ = 0⋅5 J/m2 (typical sur-
face energy), R = 8⋅31 J/mole–1 K–1, T = 453 K (typical
artificial ageing temperature) K = 2⋅65 nm, Q has been
calculated for Al–Zn–Mg and Al–Mg–Si alloys while
Q for Al–Zn–Mg alloys comes to a large value of 7⋅94,
it is a very low value of 7⋅61 × 10–7 for Al–Mg–Si
alloy.
Using the elastic strain from table 4, Q has been calcu-
lated for Al–Mg–Si and Al–Zn–Mg alloys. It comes to
2⋅9194 × 10–4 for Al–Mg–Si whereas 7⋅7 × 10–4 for
Al–Zn–Mg.
This calculation reveals that the trend of increasing Q
and Qr values from Al–Mg–Si alloy to Al–Zn–Mg alloy is
consistent with the calculation of the same quantities for
intermediate precipitates.
A reasonable assumption for cluster size would be
r = 10 nm (100 Å). The term ‘Qr’ is calculated for
Al–Zn–Mg and Al–Mg–Si alloys. Qr for Al–Zn–Mg alloy
comes to 7⋅94 nm whereas for Al–Mg–Si it is only
76⋅1 × 10–7 nm.
Now the modified stability diagram is schematically
shown in figure 5. It can be seen from the figure that
clusters are destabilized for Al–Zn–Mg alloy after two-
step ageing whereas they would be stabilized for Al–Mg–
Si alloy.
This explains how the coarse clusters are stabilized at
the artificial ageing temperature in Al–Mg–Si alloys and
thus results in lower mechanical properties as observed.
However in the case of Al–Zn–Mg alloys the clusters are
destabilized however large they are because of large Qr
and this results in fine and uniform precipitation of η′ at
the artificial ageing temperature leading to insignificant
drop in mechanical properties.
5. Conclusions
(I) A term to consider the effects of coherency strains
between cluster and matrix has been successfully incorpo-
rated in the original Pashley’s model.
(II) Experimental results show an insignificant drop in
mechanical properties for AA 7010.
(III) The high coherency strain between the cluster and
matrix has been shown to be responsible for the reduction
of delayed ageing in AA7010 alloy. It is also argued that
low coherency strain between the cluster and matrix is the
reason for delayed ageing in Al–Mg–Si alloy.
(IV) According to the extended Pashley’s model deve-
loped in the present study, cluster stability definitely gets
affected due to elastic coherency strains in the matrix. The
prediction of the new model has been verified.
Acknowledgement
The authors wish to thank the management of Hindustan
Aeronautics Ltd., Bangalore for support.
 A model for two-step ageing
References
Cahn R W, Haasen P and Kramer E J (eds) 1991 Phase trans-
formation in materials, A comprehensive treatment (New
York: VCH Publications) Vol. 5
Fast J D 1970 Entropy (Holland: Philips Technical Library) p. 91
Ghate G P, Murthy K S S and Raman K S 1984 Aluminium 60
18
Lorimer O and Nicholson R B 1966 Acta Met. 19 1009
411
Maloney S K, Hono K, Polemar I J and Ringer S P 1999 Scr.
Met. 41 1031
Murali S, Kashyap K T, Raman K S and Murthy K S S 1993
Scr. Met. 29 1421
Murayama M and Hono K 1999 Acta Met. (in press)
Pashley D W, Rhodes J W and Sendorek A 1966 J. Inst. Met. 94 41
Pashley D W, Jacobs M A and Vietz J T 1967 Philos. Mag. 16 51
Porter D A and Easterling K E 1992 Phase transformations in
metals and alloys (UK: Chapman & Hall) p. 46